What is AI good for? Automating repetitive tasks for busy people running small businesses, Berlin-based startup reckons Synthflowwhich announces a $7.4 million seed round for its no-code, SME-focused platform for AI voice assistance.
Since its founding around the spring of last year, the startup has now raised a total of $9.1 million, underscoring investors’ continued enthusiasm for accelerating genetic AI applications.
The startup also claims it’s nearing 1,000 customers — posting “double-digit” monthly growth rates since emerging from stealth development to launch its “no-code” browser-based tool in December 2023. That suggests there’s a healthy appetite. among SMEs to adopt — or at least experiment with — productive AI tools that promise easily accessible productivity gains.
The new funding will go towards R&D, according to Synthflow CEO and co-founder Hakob Astabatsyan, who says the team wants to continue building on its early momentum by increasing product utility and expanding the range of SMEs it is in. attractive.
“We have a lot of ideas. We know exactly what customers need,” he tells TechCrunch.
Astabachian, a serial entrepreneur with a business background, is a former Rocket Internet. Joining him in his latest venture is his brother, Albert, who also partnered with him in a previous no-code startup. and Sassun Mirzakhan-Saky, who brings a software engineering background and CTO expertise to the team.
While Synthflow’s product started with English language call handling because its largest markets are English speaking, it has since added German and French language versions (note: these are still in beta). Thus, its focus on the latter markets in Europe is also in focus.
End-to-end experience
Call centers were early adopters of AI voice agents, using the APIs of large language models (LLM) to power systems that could answer phone calls in a human-like way — just with tireless energy and enthusiasm 24/7, if not always perfect understanding.
Synthflow takes the idea in a slightly different direction, directly targeting service industry-focused SMEs, including those at the smaller end of the spectrum with a “no-code” DIY offering. The goal is to offer SMEs an “end-to-end” experience, according to Astabsyan, who argues that the return on investment from being able to automate basic tasks such as appointment scheduling will be immediately apparent to target businesses that they have limited resources.
“Artificial intelligence can do it in a more accessible way, more reliably, and people can do other things,” is his succinct pitch for voice assistance.
He gives the example of a handyman or mechanic who usually answered the phone themselves when not working on a job — meaning they inevitably end up missing a lot of calls and losing some work as a result. or a dentist who employs a receptionist who works limited hours so, again, isn’t always around to pick up the phone.
Having a tool that can handle basic customer questions could be a game-changer for small businesses, Astabsyan argues.
Synthflow’s aim to be SMEs necessarily means that a key focus for the startup is making AI technology accessible to non-technical users — which is why it’s created a code-free interface for its customers to design voice agents that fit their needs of their business.
“We wanted to try to make something simple,” he explains. “A layer without code on top [of AI agents] so that … business owners, business-minded people, can go and play with it and get familiar and explore what LLMs can do for their businesses.”
Synthflow’s interface allows customers to drag and drop elements to configure AI voices that can perform specific tasks for them — like scheduling meetings. Running through FAQs. or performing “information mining,” such as obtaining personal information from a potential customer so that a human can call them.
“Let’s say if someone needs to make a phone call and there’s a specific set of questions that need to be asked and specific pieces of information that need to be collected—specifically static, like address, home, etc. – the artificial intelligence is very good,” he argues. .
The customer can choose to configure the AI assistant to reveal that it is a robot. “I think it’s a very good practice to disclose that it’s a virtual assistant,” says Astabsyan. “My personal favorite opening is, ‘Hi. My name is [so-and-so], at this time, all our lines are busy. I apologize for that. I am the virtual assistant here at [the name of the business]. How can I help you;’.”
Another big utility for voice AIs is recognizing when a call should be transferred to a human agent, according to Astabsyan. So, in essence, using AI to filter incoming calls based on complexity — with automation taking care of simple requests that then leverages the benefit by freeing up human agents to spend more time on more complex customer questions.
He emphasizes that the goal is not to replace human jobs, but rather suggests that AI can help SMEs be more productive and efficient than they could otherwise be with their limited resources.
That’s why, in addition to allowing customers to deploy voice agents, Synthflow’s system is designed to take care of post-call data entry tasks — for example, adding appointments to a calendar tool. Building integrations with third-party software is therefore another big focus for the team.
“That’s what AI is so good at,” he argues. “Because he can get that information [extracted from a call] and, say, update certain fields, specifically CRM — and if you do these things at scale, across hundreds or 1,000s of calls, suddenly we see this technology advantage that we saw [when businesses first adopted] Computers.”
For voice agents, the startup is based on OpenAI’s GPT LLM, but also integrates its own AI models on top — which Astabsyan says are trained on its own data and tailored to specific customer use cases.
It says it has also built its own “voice orchestration layer” which converts the customer’s speech into text that can then be fed into the AI model as a prompt, returning an automated response that the system converts from text to speech that the customer hears as a synthetic voice on the other end of the phone line.
For now, Synthflow is focused on using AI with inbound calls – which Astabsyan suggests is the low-hanging fruit for automation for under-resourced businesses. However, it hints at more sophisticated possibilities in development, with R&D fueled by the fat seed.
One thing he says they’re working on is a feature that will allow Synthflow’s voice AIs to perform what he refers to as “live actions” or “connections” — meaning that during a call the AI will be able to perform a check live inventory in warehouse. Or pull another piece of information you requested and “push it elsewhere,” as he puts it.
It also outlines a scenario where task-focused voice AI systems will be able to expand their utility collectively. They could hand over a call to other dedicated voice AIs that are trained for different tasks requested by the customer.
“The key here is to focus on who your customers are. Because depending on who you’re building it for, your product is going to be very, very, very different,” he adds.
One impact to consider is whether voice AI and voice assistance systems live up to the productivity hype — nimbly delivering on the promise of efficiently handling an entire layer of customer questions, including expertly redirecting more complex things to the right system or human for dealing with — could end up meaning the average SME finds it has far more work than it can handle.
“I think that’s an interesting question for a lot of managers and leaders to think about, right?” he replies, discussing this scenario. “Like, if there’s so much capacity — and productivity is unleashed — how do we channel that human capital into other sectors of the economy? Because I think that question hasn’t been answered yet, but it’s a very interesting question indeed.”
Synthflow’s seed funding is being led by Singular, with participation from existing investors Atlantic Labs and several investors in the artificial intelligence space, including the founders of Krisp AI.
